AI Gave Marketing Speed. That Was Never the Hard Problem.

By Shashi Bellamkonda  ·  March 30, 2026  ·  misunderstoodmarketing.com

Tom Roach, vice president of brand strategy at Jellyfish, published a column in Marketing Week today that is worth reading carefully, not because it is contrarian, but because it names a tension most marketing teams are living inside without quite articulating it. His argument: the marketing conversation around artificial intelligence has been stuck in efficiency mode, and it needs to move to effectiveness mode. From faster and cheaper, to actually driving growth.

He is right. And the implications run deeper than the framing suggests.

The Jevons paradox, applied to marketing

Roach reaches back to a 19th-century observation to make the point. In 1865, the economist William Stanley Jevons noticed that as steam engines became more efficient at burning coal, total coal consumption did not fall. It skyrocketed. Coal was now cheaper and more productive, so it became viable for uses that were previously impossible. Efficiency created abundance, and abundance unlocked entirely new categories of value.

The parallel to AI in marketing is direct. When content production gets ten times cheaper, the rational response is not to produce the same content at lower cost and pocket the savings. The rational response is to figure out what becomes possible at that price point that was not possible before. New formats, new audiences, new experiments, new bets on creative territory that was previously too expensive to test.

Most marketing teams are not there yet. They are in the cost-reduction phase, which is a reasonable starting point. The trap is treating it as the destination.

What the data actually shows

Roach is candid about what the evidence supports and what it does not. Research conducted by System1 and Jellyfish tested 18 AI-assisted video ads and found they averaged higher scores than System1's benchmark: 3.4 stars versus a 2.3-star average for traditionally produced work. That is a real result. AI-assisted production can match or beat conventional production on standard effectiveness metrics.

But Roach is careful about what that means. Those gains, he writes, mostly reflect doing the old things faster and more efficiently. They do not represent a new model of growth. Brands that only use AI to produce more content more cheaply will not fall behind on quality. They will fall behind on speed and efficiency if they do not adopt AI at all. But keeping pace is not the same as pulling ahead.

If a chief marketing officer's only contribution to the AI agenda is to drive efficiencies and cut costs, it will be a missed opportunity.

That sentence is worth sitting with. The CMO who defines success as an AI-enabled cost reduction has answered the wrong question. The question is what growth the organization is leaving on the table by not using the abundance AI creates to find new ways to compete.

Where the real data is hiding

Here is the part of Roach's column that most marketing leaders should find uncomfortable. He observes that the real case studies of AI-enhanced growth from major brands are largely not public. Either the brands are keeping competitive results confidential, or the campaigns have not been in market long enough to produce the longitudinal data that would show genuine business impact rather than short-term engagement gains.

This matters because the current public discourse is dominated by efficiency metrics. Cost per asset. Time saved. Volume of content produced. Those are real numbers and easy to report. The effectiveness data, what share of market shifted, what brand preference moved, what purchase behavior changed as a result of AI-enabled marketing, is much harder to produce and largely absent from the conversation.

That gap shapes how budgets get set and how success gets defined. If the only numbers available are efficiency numbers, efficiency becomes the standard by which AI investments get evaluated. That is a self-reinforcing trap.

The democratization angle most strategists are underweighting

Roach makes a point about smaller businesses that deserves more attention in the strategy conversation. The previous digital revolution democratized media: a local business could buy the same ad space as a global brand. What it did not democratize was creative production. Professional-quality video, sophisticated brand content, and high-production creative still required agencies and production companies that most small businesses could not afford.

Generative AI changes that. For the first time, a small business can produce professional-quality video and content without a network agency or a production house. The economic value of that shift across the broader economy is likely to be enormous, even if it generates creative output that larger brands would not be proud of. Roach suggests a more generous read: most people want to do a good job, and organizations new to a medium tend to improve over time. Television advertisers typically take three to four years to learn how to make effective television creative.

The implication for established brands is that the competitive context is shifting. The companies entering your category with AI-generated content at a fraction of your cost are not going away. Some will be bad. Some will learn quickly.

Generative Engine Optimization as a leading indicator

One area Roach flags as a genuine early example of AI enabling growth rather than just efficiency is Generative Engine Optimization, or GEO. As AI tools increasingly become the first point of contact in a buyer's research process, visibility inside those tools becomes a marketing objective in its own right. GEO is the emerging discipline of optimizing for that visibility, building on the foundations of search engine optimization but oriented toward how AI models surface and describe brands.

New research from LinkedIn reinforces why this matters in business-to-business marketing specifically. Their study finds that traditional engagement metrics like reach, clicks, and content interaction no longer reliably predict whether a brand ends up in the consideration set when a purchase decision is made. The study frames this as a "buyability" problem. Being visible is no longer enough. Being present and credible in the sources AI tools draw on when building a shortlist is the new objective.

For marketing technology buyers, this is a concrete strategy question today, not a future-state consideration.

The practitioner's lens

Having worked in marketing for twenty-five years, including in roles where the budget pressure was constant and the mandate to do more with less was permanent, I recognize the pattern Roach is describing. Every technology cycle starts with efficiency. Email replaced direct mail on cost. Paid search replaced broadcast on targeting precision. Social replaced print on reach per dollar. In each case, the first-mover advantage went to the organizations that figured out what became possible with the new technology, not just what became cheaper.

AI is no different. The efficiency gains are real and worth capturing. The harder work is asking what you can now attempt that was previously out of reach, and having the organizational confidence to pursue it before the results are guaranteed. That is the CMO's job. The tools do not make the judgment call.